Metabolic and Transcriptional Pathway Inference using State Space Models: An Application to Understanding Acid Response

Functional genomics technologies have increased our ability to accumulate information about a biological system to a degree that would have been unimaginable only a few years ago. It is in fact possible to monitor the expression of thousands of genes, proteins and metabolites in single experiments. This incredible amount of information does not however result in an automatic increase in biological knowledge. A plethora of algorithms to classify and describe the complexity of these datasets have been developed in the attempt to extract useful information. Although these methods have demonstrated to be useful they do not provide models of the underlying molecular networks that lend themselves to statistical hypothesis testing. There is therefore the need to develop robust inference techniques to generate hypothesis on the causal relationship between genes, proteins and metabolites for which we have measurements. The ability in infer causal relationships from large scale descriptive data would change our approach to biological investigation and would provide a better understanding of an organisms complex network of regulatory interactions. The project we propose is based on a computational framework for network inference recently developed by the applicants and aims to reconstruct transcriptional and metabolic networks representative of the response of E. coli to acid stress. Although global proteomics data obtained using MALDI-TOF HPLC would offer some potential advantages (or at least be complementary) over NMR spectrum based metabolomics, it is prohibitively expensive and too unwieldy to apply to the sequential approach required for iterative building and testing of mathematical models. The experimental system we have chosen (E. coli acid response) is extremely suitable to understanding the physiology of a bacterial pathogen and, importantly, it is amenable to rapid experimental verification.<P> In conclusion, this project has four strengths:<OL> <LI> A highly reproducible biological system controlled using a biological reactor (chemostat); <LI> A modelling framework that allows reverse engineering from continuous data and includes hidden factors; <LI> Highly reliable and relatively inexpensive technologies for monitoring gene expression and metabolite concentration; <LI> An extremely easy to manipulate experimental system for model verification.